LNRLMI: Linear neighbour representation for predicting lncRNA-miRNA interactions

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Long non-coding RNAs (lncRNAs) and microRNAs (miRNAs) are key regulators in the competing endogenous RNA (ceRNA) network mechanism.
• Understanding lncRNA-miRNA interactions is vital for gene regulation, complementing experimental methods like CLIP-seq.
• Developing computational tools for predicting these interactions is crucial but underexplored.

Purpose of the Study:

• To propose a novel computational approach, Linear Neighbour Representation for lncRNA-miRNA Interactions (LNRLMI), for predicting lncRNA-miRNA interactions.
• To provide an effective in silico tool for identifying potential lncRNA-miRNA interactions for experimental validation.
• To enhance the understanding of the ceRNA regulatory mechanism.

Main Methods:

• Constructed a bipartite network integrating known interaction data with expression profile similarities of lncRNAs and miRNAs.
• Applied the linear neighbour representation method to build a prediction model based on the integrated network.
• Evaluated model performance using k-fold cross-validation.

Main Results:

• LNRLMI achieved high prediction accuracy, with average AUCs of 0.8475 (2-fold), 0.8960 (5-fold), and 0.9069 (10-fold) cross-validation.
• Comparative experiments demonstrated the feasibility and effectiveness of LNRLMI against other methods.
• The approach successfully integrated diverse side information for accurate predictions.

Conclusions:

• LNRLMI is a feasible and effective computational tool for predicting lncRNA-miRNA interactions.
• The method's performance highlights the utility of integrating network and expression data.
• LNRLMI is anticipated to be valuable for predicting non-coding RNA regulatory networks involving lncRNAs and miRNAs.